Respiratory Liver Motion Extraction for Proton Therapy with Active Scanning Beam Delivery by Deformable Registration
The Abramson Cancer Center of the University of Pennsylvania
Last Modified: May 12, 2011
Presenter: Ye Zhang1,2, Antje-Christin Knopf1,Dirk Boye1, Tony Lomax1
Presenter's Institution: Center for Proton Therapy, Paul Scherrer Institut, Switzerland and Computer Vision Laboratory, ETH Zurich, Switzerland
- Although proton therapy with active scanning beam delivery has advantages to conventional radiation therapy when treating cancer in a location that remains relatively motionless, treating mobile targets that move during treatment, particularly the lung and the abdomen, remains a challenge.
- Intrafraction liver motion can dramatically effect the dose distribution and reliable motion estimations are critical due to the sharp distal margin of the proton beam.
- Active beam scanning is sensitive to motion and density variations, and estimations of motion using a blurring technique can create hot and cold spots within the target.
- In order to optimize treatment of targets, or cancers, in locations of the body such as abdomen and liver, where motion can be significant, it is necessary to analyze the dose distribution with respect to time.
- The purpose of the study was to evaluate the variability in two deformable motion registration models and attempt to minimize the subsequent dose variation using multiple fields and re-scanning.
- Liver motion secondary to respirations was extracted from 4D CTs (10 phases) via deformable registration models, B-Splines and Demons.
- The accuracy of these deformable registration models was compared to traditional contour based tracking with landmark-based tracking.
- To estimate liver motion as a dense displacement vector, the water equivalent depth change was calculated.
- Subsequently, the treatment plans for the three liver patient cases were compared using 3D dose calculations on deformed dose grids with density changes to evaluate the impact of the deformable registration models.
- The different motion estimations were significantly different for individual registration models with a maximum difference of 26.8% and a mean difference of 3.3%.
- While the relative error for motion estimates was similar for the two deformable registration models, the displacement was different.
- Treatment plans created using only a few fields and no re-scanning had significant differences in the resulting 4D dose calculations.
- To solve the problem of varying motion estimates based on the individual algorithm used to calculate motion, the authors evaluated the use of multiple fields and re-scanning to minimize the difference.
- If re-scanning was used and multiple fields (three in this study) were employed, the dose calculations proved to be small, diminishing the influence of different motion estimates.
- Due to the difficult nature of deformable registration of the liver with 4DCT, different registration methods to evaluate respiratory liver motion can provide significantly different motion assessments due to registration errors and registration bias.
- This mean difference could have clinical implications and should be compensated for in some manner.
- In order to accurately calculate dose distributions for scanned proton therapy, error bars should be employed to echo uncertainties.
- Preliminary results reveal rescanning is an effective approach for managing target motion due to its ability to smear the interplay with motions.
- Similarly, increasing the number of fields used in treatment planning can aid in diminishing the effect of target motion.
- Finally, evaluation of liver motion with 4D-MRI may prove beneficial in this area.
- As proton beam therapy becomes more widespread and its use is broadened across various treatment sites, the impact of organ motion must be taken into consideration for treatment planning.
- Employment of methods to understand and account for organ motion is paramount to accurately predict intrafraction organ motion with the use of proton therapy
- Certainly, clinical testing of prototype designs is essential to ensure the accurate implementation of treatments prior to widespread use.